A Compact and Infrastructure-free Confined Space Sensor for 3D Scanning and SLAM

Abstract

High-precision inspection and metrology in short-range and tight spaces are challenging due to the lack of a commercial off-the-shelf (COTS) 3D scanner that is compact and does not rely on any external infrastructure (e.g., fiducial markers, motion-capture cameras, or laser tracking interferometer) to provide positioning or localization support. This paper presents a hardware and software design framework for creating a low-cost, miniature, yet intelligent sensor that is able to capture visual imagery, reconstruct 3D geometry, and most importantly, perform Simultaneous Localization and Mapping (SLAM) without sensory feedback from external devices. We further present an example sensor design that follows this framework, with experiments to evaluate and compare the sensor performance against a state-of-the-art COTS sensor, under both feature-sparse and feature-dense scenarios to simulate industrial and household 3D scanning application scenarios.